Self- and semi-supervised learning methods have been actively investigated to reduce labeled training data or enhance the model performance. However, the approach mostly focus on in-domain performance for public datasets. In this study, we utilize the combination of self- and semi-supervised learning methods to solve unseen domain adaptation problem in a large-scale production setting for online ASR model. This approach demonstrates that using the source domain data with a small fraction of the target domain data (3%) can recover the performance gap compared to a full data baseline: relative 13.5% WER improvement for target domain data.
@article{arxiv.2110.00165,
title = {Large-scale ASR Domain Adaptation using Self- and Semi-supervised Learning},
author = {Dongseong Hwang and Ananya Misra and Zhouyuan Huo and Nikhil Siddhartha and Shefali Garg and David Qiu and Khe Chai Sim and Trevor Strohman and Françoise Beaufays and Yanzhang He},
journal= {arXiv preprint arXiv:2110.00165},
year = {2022}
}